On Mon, 2005-09-12 at 20:42 +0100, Mark Adams wrote: > FYI, I've just downloaded the interlace tests you refer to from > http://www.kingcot.eclipse.co.uk/unichrome/tvoutTest.html. Here's > what happens on my machine (VIA EPIA MII-12000, 512 MB RAM, Fedora > Core 3, various patches): > > df_xine -l0 interlace_test2.mpg - picture entirely as described on the > web site except that the field parity drifts in and out such that the > white bar is sometimes moving very smoothly and sometimes a bit > juddery; 60% CPU usage (RGB16) > > df_xine -s -l0 interlace_test2.mpg - picture entirely as described on > the web site with the field parity remaining constant but being > randomly correct or incorrect; 60% CPU usage (RGB16) > > df_xine -s -l1 interlace_test2.mpg - black screen > > export DFB_CLE266_UNDERLAY=1 > df_xine -s -l1 interlace_test2.mpg - picture entirely as described on > the web site but always showing incorrect field parity; 36% CPU usage > (YV12) > > df_xine -s -l1 -p rgb16 interlace_test2.mpg - picture entirely as > described on the web site but always showing incorrect field parity; > 60% CPU usage (RGB16) > > df_xine -s -l1 tv.mpg (some captured broadcast TV) - always correct > field parity; 52% CPU usage (YV12) > > Conclusions: > > The file interlace_test2.mpg has _incorrect_ field parity (at least > it's opposite to broadcast TV). This I have since confirmed by > looking at a decoded frame: field 2 is clearly on the even lines (0, > 2, 4 etc.). > > My patches only alter layer _1_ to provide correct frame > synchronisation (normally you wouldn't be displaying interlaced > material on layer 0 so this seems reasonable) > > In all cases, my machine copes fine with the workload. > > You say you can't get interlace_test2.mpg to display at all with > df_xine which suggests you still have some fundamental problems to > sort out.
Thanks, Mark. This gives me a reference point: your hardware is identical to mine so all I need to do now is work out why mine does weird things! :-s I'm back after a week's holiday so I may as well start again from scratch with this rather than try to work out where I had got to! Are you using cvs or release DirectFB? The fact that I get what looks like the same jerky playback from both df_xine and vdr with softdevice makes me think that the problem lies somewhere with DirectFB rather than with xine. Do you know if there are any kernel options to avoid such as preemt or low-latency type of stuff? One thought I've had (but not had a chance to test yet) is some bizarre setting in ~/.xine/config but that wouldn't have any effect on softdevice! That _may_ explain why it wouldn't play the interlace_test2.mpg file properly for me. Any thoughts on what to test? Cheers, Laz _______________________________________________ directfb-users mailing list [email protected] http://mail.directfb.org/cgi-bin/mailman/listinfo/directfb-users
